8 Zooarchaeology

Page under construction

8.1 Case studies

The following map shows the sites under investigation, divided by chronology. Please select the desired chronology (or chronologies) from the legend on the right.

Legend: R = Roman, LR = Late Roman, EMA = Early Middle Ages, Ma = 11th c. onwards

8.2 Medians

The faunal dataset is large (434+ records) and diversified. Looking at the distributions of each animal, the curve is not gaussian. The best choice for non-normal curves is to use medians instead of means to come up with figures that are less dependent on outliers. The function Medians_Chrono_Zoo() (Section 3.1) can be used to return as output weighted medians for each chronology. The in-depth description of how weights are calculated for each sample can be found in Section 6.4.1. To summarise, sites with a very large (i.e. fuzzy) chronology contribute less to the calculation of the median. Table 8.1 provides the median values of the main categories of faunal remains for each chronology, and Table 8.2 the median values for each century. Stronger colours in the cells indicate higher values.

Show the code

Medians_Categorised_per_Chronology_ZOO <- 
  data.frame(
    Medians_Chrono_Zoo(zooarch_cond, "R")*100,
    Medians_Chrono_Zoo(zooarch_cond, "LR")*100,
    Medians_Chrono_Zoo(zooarch_cond, "EMA")*100,
    Medians_Chrono_Zoo(zooarch_cond, "Ma")*100
  )

# Round to 2 digits
Medians_Categorised_per_Chronology_ZOO <- round(Medians_Categorised_per_Chronology_ZOO, 2)

## Weighted medians per century ##
Medians_ZOO_Centuries <- data.frame(
  "I BCE" = zooarch_tables(zooarch_cond, -1)$Medians,  
  "I CE" = zooarch_tables(zooarch_cond, 1)$Medians,
  "II CE" = zooarch_tables(zooarch_cond, 2)$Medians,
  "III CE" = zooarch_tables(zooarch_cond, 3)$Medians,
  "IV CE" = zooarch_tables(zooarch_cond, 4)$Medians,
  "V CE" = zooarch_tables(zooarch_cond, 5)$Medians,
  "VI CE" = zooarch_tables(zooarch_cond, 6)$Medians,
  "VII CE" = zooarch_tables(zooarch_cond, 7)$Medians,
  "VIII CE" = zooarch_tables(zooarch_cond, 8)$Medians,
  "IX CE" = zooarch_tables(zooarch_cond, 9)$Medians,
  "X CE" = zooarch_tables(zooarch_cond, 10)$Medians,
  "XI CE" = zooarch_tables(zooarch_cond, 11)$Medians
)

# Assigning the colnames (optional - instead of roman numerals)
colnames(Medians_ZOO_Centuries) <- c("1st c. BCE", "1st c. CE", "2nd c.", "3rd c.", "4th c.", "5th c.", "6th c.", "7th c.", "8th c.", "9th c.", "10th c.", "11th c.")

# Rounding the medians
Medians_ZOO_Centuries <- round(Medians_ZOO_Centuries, digits=2)

# Removing categories that are not necessary
Medians_ZOO_Centuries <- Medians_ZOO_Centuries[-c(6:9),]

Table 8.1: Weighted medians of zooarchaeological remains, divided by chronology.
	Chronologies
	R	LR	EMA	Ma
Pigs	48.0	38.62	35.00	35.62
Cattle	10.1	8.36	18.00	19.00
Caprine	25.0	22.08	29.00	27.00
Dom..Fowl	4.0	6.00	5.00	5.00
Edible.W..Mammals	5.0	3.00	3.00	3.00
Fish	1.0	2.00	3.93	1.00
Mollusca	11.0	8.00	4.00	2.89
Unedible.Dom..Mammals	2.0	3.00	3.95	2.00
Unedible.Wild.Mammals	1.0	1.00	1.00	1.00

Pigs’ medians from the Italian peninsula are the highest in each chronology, although their values decrease after the Roman age peak. Cattle medians slightly decrease after the Roman age, even though surprisingly (put a reference here to literature review to explain why surprisingly) the values increase again (18–19.71%) during the early Medieval and Medieval age. The trends for sheeps and goats are also interesting. During the Roman age the Italian median is 25%, slightly decreasing in the 3^rd to the 5^th century, and increasing again after. When discussing sheep-farming, one must always consider the geographical features from which the data is being collected. This will be discussed later on in the chapter, where more regional and geographical trends will be provided. Domestic fowl (chickens and geese) has quite stable values of 4-5%, with a peak of 7.68% in the 11^th century. Wild game peaks during the Roman age, with a median value of 5%, reaching a minimum in the early Middle ages (2%) and rising again in the 11^th century. Two considerations must be made for game consumption. The first is that as we will see later on, game consumption is strongly related to the site typology. Secondly, the Roman age value is pulled up by assemblages from the 1^st century BCE. After that, the values strongly decrease and by looking at the individual centuries the medians from the 7^th century onwards are much higher (ranging from 1.42% to 2.09%).

Table 8.2: Weighted medians of zooarchaeological remains, divided by century.
	Faunal remains
	Pigs	Cattle	Caprine	Domestic fowl	Wild game
1st c. BCE	39.70	11.41	26.07	0.00	0.76
1st c. CE	40.89	11.31	25.48	0.00	0.58
2nd c.	48.20	10.28	22.33	0.79	0.00
3rd c.	41.57	8.24	19.22	1.53	0.73
4th c.	34.15	9.56	23.23	1.79	0.98
5th c.	34.01	13.38	24.69	2.33	0.88
6th c.	31.83	20.66	28.55	2.75	0.65
7th c.	30.42	18.52	30.14	4.51	1.58
8th c.	33.63	13.89	30.13	2.70	1.18
9th c.	37.54	11.76	23.32	1.16	1.54
10th c.	35.77	14.36	25.64	1.63	1.93
11th c.	34.08	19.34	27.44	1.91	2.09
^* The color gradients in this table are used to indicate the chronologies.

8.2.1 Medians of faunal remains by context type

The weighted medians included below have been generated using the package dplyr and the summarize() function, applied to the exported relative proportion table (using the custom function zooarch_tables_general(), described in Section 3.2). The medians have been calculated for four animal categories (pigs, cattle, caprine, and game) for each site type and chronology. After, similar context types have been merged to simplify the reading; for example, the category Castle has been merged with the category Castrum, as they both indicate élite/military fortified contexts.

8.2.2 Medians of faunal remains by macro region

The process for generating weighted medians for the three Italian macro regions (Southern, Central and Northern Italy) has followed the same logic used in the previous section. The medians have been calculated for four animal categories (pigs, cattle, caprine, and game) for each macro area and chronology.

8.2.3 Medians of faunal remains by geography type

Weighted medians have been generated for the four geographies considered (plain, hill, hilltop, coast, and mountain), after the categories Hill and Hilltop have been merged. The medians have been calculated for four animal categories (pigs, cattle, caprine, and game) for each geography and chronology.

8.2.4 Caprine vs altitude

9 In progress

9.1 Test 1: GLM-Pigs %

Question: Is the animal X - let’s say Pigs - more strongly associated to a particular settlement type during a certain chronology?

Expectation 1: Pigs % should be higher in urban and fortified settlements, as they are animals which are only used for meat and can sustain large populations or the military.

Expectation 2: Possibly pigs would increase in villas in the Late Roman age, as their production shifts to a more extensive agriculture (Source: Historical literature).

Expectation 3: If urban density decreases during the late Roman and early Medieval phase, but increases in the Medieval age, do the pigs % follow similar trends?

Note however that there is a temporal divergence in the means of pigs for Central-Southern Italy (Pigs % decrease from the Roman to the Medieval age) and Northern Italy (Pigs % increase from the Roman to the Medieval age).

Before running the GLM, a quick look at the distribution of Pigs in each chronology and site type.

[1] "Number of samples:"


  R  LR EMA  Ma 
169 174 130  60

[1] "Chronological means and sd:"

# A tibble: 4 × 3
  Chronology  mean    SD
  <fct>      <dbl> <dbl>
1 R           0.43  0.22
2 LR          0.36  0.19
3 EMA         0.33  0.16
4 Ma          0.37  0.15

Plots showing very high variability in each site type:

We can now run the GLM:

m_zoo_st <- glm(Pigs~Chronology*Type, 
               # weights = Tot_NISP, 
                family = "quasibinomial",
                data=Animals_Df
                )

# This model uses % data, so I am not sure I should use the weights.

Somehow the results for Ma:Urban are NA, even though the category is not empty (as is the case for Rural site, villas in the Middle ages) and there are no NAs.


Call:
glm(formula = Pigs ~ Chronology * Type, family = "quasibinomial", 
    data = Animals_Df)

Deviance Residuals: 
     Min        1Q    Median        3Q       Max  
-1.07441  -0.26547   0.01331   0.24466   1.06624  

Coefficients: (2 not defined because of singularities)
                                    Estimate Std. Error t value Pr(>|t|)    
(Intercept)                           0.1474     0.2335   0.631   0.5281    
ChronologyLR                         -0.6914     0.3715  -1.861   0.0632 .  
ChronologyEMA                        -0.4723     0.2870  -1.645   0.1005    
ChronologyMa                         -0.5183     0.1579  -3.283   0.0011 ** 
TypeRural                            -1.0929     0.2630  -4.156 3.79e-05 ***
TypeRural site, villa                -0.4605     0.2805  -1.641   0.1013    
TypeUrban                            -0.1778     0.2215  -0.802   0.4227    
ChronologyLR:TypeRural                0.6469     0.4090   1.582   0.1143    
ChronologyEMA:TypeRural               0.5664     0.3416   1.658   0.0979 .  
ChronologyMa:TypeRural                0.6182     0.3136   1.972   0.0492 *  
ChronologyLR:TypeRural site, villa    0.8783     0.4247   2.068   0.0391 *  
ChronologyEMA:TypeRural site, villa   0.4242     0.4110   1.032   0.3025    
ChronologyMa:TypeRural site, villa        NA         NA      NA       NA    
ChronologyLR:TypeUrban                0.2011     0.3728   0.539   0.5899    
ChronologyEMA:TypeUrban              -0.2711     0.2936  -0.923   0.3562    
ChronologyMa:TypeUrban                    NA         NA      NA       NA    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

(Dispersion parameter for quasibinomial family taken to be 0.1357735)

    Null deviance: 91.342  on 532  degrees of freedom
Residual deviance: 77.731  on 519  degrees of freedom
AIC: NA

Number of Fisher Scoring iterations: 4

Pseudo R-squared:  0.149007

Note:
  2 values in the Chronology*Type effect are not estimable

9.2 Test 2: GLMM - Pigs %

Not working anymore somehow!

# Load brms
library(brms)

# Should probably using binomial, but it giving weird results and a warning message
# about specifying trials on the left side of the formula
pigs_glm_mod <- brm(Pigs|trials(Tot_NISP) ~ 1 +Type+Chronology + (1|ID),
                    data = Animals_Df,
                    beta_binomial(),
                    iter=3000,
                    warmup = 500,
                    chains = 2, 
                    cores = 4,
                    seed=123
                    )

Besides the problem of the ID column being informative or the correct one to use, how do I interpret the results? Are the estimates estimates of the mean that I have to add up to the intercept? Is this true also for the interaction estimates?

##Test 3 - Betabin